ccid3.c 36 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285
  1. /*
  2. * net/dccp/ccids/ccid3.c
  3. *
  4. * Copyright (c) 2005 The University of Waikato, Hamilton, New Zealand.
  5. * Copyright (c) 2005-6 Ian McDonald <ian.mcdonald@jandi.co.nz>
  6. *
  7. * An implementation of the DCCP protocol
  8. *
  9. * This code has been developed by the University of Waikato WAND
  10. * research group. For further information please see http://www.wand.net.nz/
  11. *
  12. * This code also uses code from Lulea University, rereleased as GPL by its
  13. * authors:
  14. * Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon
  15. *
  16. * Changes to meet Linux coding standards, to make it meet latest ccid3 draft
  17. * and to make it work as a loadable module in the DCCP stack written by
  18. * Arnaldo Carvalho de Melo <acme@conectiva.com.br>.
  19. *
  20. * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
  21. *
  22. * This program is free software; you can redistribute it and/or modify
  23. * it under the terms of the GNU General Public License as published by
  24. * the Free Software Foundation; either version 2 of the License, or
  25. * (at your option) any later version.
  26. *
  27. * This program is distributed in the hope that it will be useful,
  28. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  29. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  30. * GNU General Public License for more details.
  31. *
  32. * You should have received a copy of the GNU General Public License
  33. * along with this program; if not, write to the Free Software
  34. * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  35. */
  36. #include "../ccid.h"
  37. #include "../dccp.h"
  38. #include "lib/packet_history.h"
  39. #include "lib/loss_interval.h"
  40. #include "lib/tfrc.h"
  41. #include "ccid3.h"
  42. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  43. static int ccid3_debug;
  44. #define ccid3_pr_debug(format, a...) DCCP_PR_DEBUG(ccid3_debug, format, ##a)
  45. #else
  46. #define ccid3_pr_debug(format, a...)
  47. #endif
  48. static struct dccp_tx_hist *ccid3_tx_hist;
  49. static struct dccp_rx_hist *ccid3_rx_hist;
  50. static struct dccp_li_hist *ccid3_li_hist;
  51. /*
  52. * Transmitter Half-Connection Routines
  53. */
  54. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  55. static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
  56. {
  57. static char *ccid3_state_names[] = {
  58. [TFRC_SSTATE_NO_SENT] = "NO_SENT",
  59. [TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
  60. [TFRC_SSTATE_FBACK] = "FBACK",
  61. [TFRC_SSTATE_TERM] = "TERM",
  62. };
  63. return ccid3_state_names[state];
  64. }
  65. #endif
  66. static void ccid3_hc_tx_set_state(struct sock *sk,
  67. enum ccid3_hc_tx_states state)
  68. {
  69. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  70. enum ccid3_hc_tx_states oldstate = hctx->ccid3hctx_state;
  71. ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
  72. dccp_role(sk), sk, ccid3_tx_state_name(oldstate),
  73. ccid3_tx_state_name(state));
  74. WARN_ON(state == oldstate);
  75. hctx->ccid3hctx_state = state;
  76. }
  77. /*
  78. * Compute the initial sending rate X_init according to RFC 3390:
  79. * w_init = min(4 * MSS, max(2 * MSS, 4380 bytes))
  80. * X_init = w_init / RTT
  81. * For consistency with other parts of the code, X_init is scaled by 2^6.
  82. */
  83. static inline u64 rfc3390_initial_rate(struct sock *sk)
  84. {
  85. const struct dccp_sock *dp = dccp_sk(sk);
  86. const __u32 w_init = min(4 * dp->dccps_mss_cache,
  87. max(2 * dp->dccps_mss_cache, 4380U));
  88. return scaled_div(w_init << 6, ccid3_hc_tx_sk(sk)->ccid3hctx_rtt);
  89. }
  90. /*
  91. * Recalculate t_ipi and delta (should be called whenever X changes)
  92. */
  93. static inline void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
  94. {
  95. /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
  96. hctx->ccid3hctx_t_ipi = scaled_div32(((u64)hctx->ccid3hctx_s) << 6,
  97. hctx->ccid3hctx_x);
  98. /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
  99. hctx->ccid3hctx_delta = min_t(u32, hctx->ccid3hctx_t_ipi / 2,
  100. TFRC_OPSYS_HALF_TIME_GRAN);
  101. ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
  102. hctx->ccid3hctx_t_ipi, hctx->ccid3hctx_delta,
  103. hctx->ccid3hctx_s, (unsigned)(hctx->ccid3hctx_x >> 6));
  104. }
  105. /*
  106. * Update X by
  107. * If (p > 0)
  108. * X_calc = calcX(s, R, p);
  109. * X = max(min(X_calc, 2 * X_recv), s / t_mbi);
  110. * Else
  111. * If (now - tld >= R)
  112. * X = max(min(2 * X, 2 * X_recv), s / R);
  113. * tld = now;
  114. *
  115. * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
  116. * fine-grained resolution of sending rates. This requires scaling by 2^6
  117. * throughout the code. Only X_calc is unscaled (in bytes/second).
  118. *
  119. */
  120. static void ccid3_hc_tx_update_x(struct sock *sk, struct timeval *now)
  121. {
  122. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  123. __u64 min_rate = 2 * hctx->ccid3hctx_x_recv;
  124. const __u64 old_x = hctx->ccid3hctx_x;
  125. /*
  126. * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
  127. * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
  128. * For consistency with X and X_recv, min_rate is also scaled by 2^6.
  129. */
  130. if (unlikely(hctx->ccid3hctx_idle)) {
  131. min_rate = rfc3390_initial_rate(sk);
  132. min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
  133. }
  134. if (hctx->ccid3hctx_p > 0) {
  135. hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
  136. min_rate);
  137. hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
  138. (((__u64)hctx->ccid3hctx_s) << 6) /
  139. TFRC_T_MBI);
  140. } else if (timeval_delta(now, &hctx->ccid3hctx_t_ld) -
  141. (suseconds_t)hctx->ccid3hctx_rtt >= 0) {
  142. hctx->ccid3hctx_x =
  143. max(min(2 * hctx->ccid3hctx_x, min_rate),
  144. scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
  145. hctx->ccid3hctx_rtt));
  146. hctx->ccid3hctx_t_ld = *now;
  147. }
  148. if (hctx->ccid3hctx_x != old_x) {
  149. ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
  150. "X_recv=%u\n", (unsigned)(old_x >> 6),
  151. (unsigned)(hctx->ccid3hctx_x >> 6),
  152. hctx->ccid3hctx_x_calc,
  153. (unsigned)(hctx->ccid3hctx_x_recv >> 6));
  154. ccid3_update_send_interval(hctx);
  155. }
  156. }
  157. /*
  158. * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
  159. * @len: DCCP packet payload size in bytes
  160. */
  161. static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
  162. {
  163. const u16 old_s = hctx->ccid3hctx_s;
  164. hctx->ccid3hctx_s = old_s == 0 ? len : (9 * old_s + len) / 10;
  165. if (hctx->ccid3hctx_s != old_s)
  166. ccid3_update_send_interval(hctx);
  167. }
  168. /*
  169. * Update Window Counter using the algorithm from [RFC 4342, 8.1].
  170. * The algorithm is not applicable if RTT < 4 microseconds.
  171. */
  172. static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock *hctx,
  173. struct timeval *now)
  174. {
  175. suseconds_t delta;
  176. u32 quarter_rtts;
  177. if (unlikely(hctx->ccid3hctx_rtt < 4)) /* avoid divide-by-zero */
  178. return;
  179. delta = timeval_delta(now, &hctx->ccid3hctx_t_last_win_count);
  180. DCCP_BUG_ON(delta < 0);
  181. quarter_rtts = (u32)delta / (hctx->ccid3hctx_rtt / 4);
  182. if (quarter_rtts > 0) {
  183. hctx->ccid3hctx_t_last_win_count = *now;
  184. hctx->ccid3hctx_last_win_count += min_t(u32, quarter_rtts, 5);
  185. hctx->ccid3hctx_last_win_count &= 0xF; /* mod 16 */
  186. ccid3_pr_debug("now at %#X\n", hctx->ccid3hctx_last_win_count);
  187. }
  188. }
  189. static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
  190. {
  191. struct sock *sk = (struct sock *)data;
  192. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  193. struct timeval now;
  194. unsigned long t_nfb = USEC_PER_SEC / 5;
  195. bh_lock_sock(sk);
  196. if (sock_owned_by_user(sk)) {
  197. /* Try again later. */
  198. /* XXX: set some sensible MIB */
  199. goto restart_timer;
  200. }
  201. ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
  202. ccid3_tx_state_name(hctx->ccid3hctx_state));
  203. hctx->ccid3hctx_idle = 1;
  204. switch (hctx->ccid3hctx_state) {
  205. case TFRC_SSTATE_NO_FBACK:
  206. /* RFC 3448, 4.4: Halve send rate directly */
  207. hctx->ccid3hctx_x = max(hctx->ccid3hctx_x / 2,
  208. (((__u64)hctx->ccid3hctx_s) << 6) /
  209. TFRC_T_MBI);
  210. ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
  211. "bytes/s\n", dccp_role(sk), sk,
  212. ccid3_tx_state_name(hctx->ccid3hctx_state),
  213. (unsigned)(hctx->ccid3hctx_x >> 6));
  214. /* The value of R is still undefined and so we can not recompute
  215. * the timout value. Keep initial value as per [RFC 4342, 5]. */
  216. t_nfb = TFRC_INITIAL_TIMEOUT;
  217. ccid3_update_send_interval(hctx);
  218. break;
  219. case TFRC_SSTATE_FBACK:
  220. /*
  221. * Modify the cached value of X_recv [RFC 3448, 4.4]
  222. *
  223. * If (p == 0 || X_calc > 2 * X_recv)
  224. * X_recv = max(X_recv / 2, s / (2 * t_mbi));
  225. * Else
  226. * X_recv = X_calc / 4;
  227. *
  228. * Note that X_recv is scaled by 2^6 while X_calc is not
  229. */
  230. BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
  231. if (hctx->ccid3hctx_p == 0 ||
  232. (hctx->ccid3hctx_x_calc > (hctx->ccid3hctx_x_recv >> 5))) {
  233. hctx->ccid3hctx_x_recv =
  234. max(hctx->ccid3hctx_x_recv / 2,
  235. (((__u64)hctx->ccid3hctx_s) << 6) /
  236. (2 * TFRC_T_MBI));
  237. if (hctx->ccid3hctx_p == 0)
  238. dccp_timestamp(sk, &now);
  239. } else {
  240. hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
  241. hctx->ccid3hctx_x_recv <<= 4;
  242. }
  243. /* Now recalculate X [RFC 3448, 4.3, step (4)] */
  244. ccid3_hc_tx_update_x(sk, &now);
  245. /*
  246. * Schedule no feedback timer to expire in
  247. * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
  248. * See comments in packet_recv() regarding the value of t_RTO.
  249. */
  250. t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
  251. break;
  252. case TFRC_SSTATE_NO_SENT:
  253. DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk), sk);
  254. /* fall through */
  255. case TFRC_SSTATE_TERM:
  256. goto out;
  257. }
  258. restart_timer:
  259. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  260. jiffies + usecs_to_jiffies(t_nfb));
  261. out:
  262. bh_unlock_sock(sk);
  263. sock_put(sk);
  264. }
  265. /*
  266. * returns
  267. * > 0: delay (in msecs) that should pass before actually sending
  268. * = 0: can send immediately
  269. * < 0: error condition; do not send packet
  270. */
  271. static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
  272. {
  273. struct dccp_sock *dp = dccp_sk(sk);
  274. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  275. struct timeval now;
  276. suseconds_t delay;
  277. BUG_ON(hctx == NULL);
  278. /*
  279. * This function is called only for Data and DataAck packets. Sending
  280. * zero-sized Data(Ack)s is theoretically possible, but for congestion
  281. * control this case is pathological - ignore it.
  282. */
  283. if (unlikely(skb->len == 0))
  284. return -EBADMSG;
  285. dccp_timestamp(sk, &now);
  286. switch (hctx->ccid3hctx_state) {
  287. case TFRC_SSTATE_NO_SENT:
  288. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  289. (jiffies +
  290. usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
  291. hctx->ccid3hctx_last_win_count = 0;
  292. hctx->ccid3hctx_t_last_win_count = now;
  293. /* Set t_0 for initial packet */
  294. hctx->ccid3hctx_t_nom = now;
  295. hctx->ccid3hctx_s = skb->len;
  296. /*
  297. * Use initial RTT sample when available: recommended by erratum
  298. * to RFC 4342. This implements the initialisation procedure of
  299. * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
  300. */
  301. if (dp->dccps_syn_rtt) {
  302. ccid3_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt);
  303. hctx->ccid3hctx_rtt = dp->dccps_syn_rtt;
  304. hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
  305. hctx->ccid3hctx_t_ld = now;
  306. } else {
  307. /* Sender does not have RTT sample: X = MSS/second */
  308. hctx->ccid3hctx_x = dp->dccps_mss_cache;
  309. hctx->ccid3hctx_x <<= 6;
  310. }
  311. ccid3_update_send_interval(hctx);
  312. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
  313. break;
  314. case TFRC_SSTATE_NO_FBACK:
  315. case TFRC_SSTATE_FBACK:
  316. delay = timeval_delta(&hctx->ccid3hctx_t_nom, &now);
  317. ccid3_pr_debug("delay=%ld\n", (long)delay);
  318. /*
  319. * Scheduling of packet transmissions [RFC 3448, 4.6]
  320. *
  321. * if (t_now > t_nom - delta)
  322. * // send the packet now
  323. * else
  324. * // send the packet in (t_nom - t_now) milliseconds.
  325. */
  326. if (delay - (suseconds_t)hctx->ccid3hctx_delta >= 0)
  327. return delay / 1000L;
  328. ccid3_hc_tx_update_win_count(hctx, &now);
  329. break;
  330. case TFRC_SSTATE_TERM:
  331. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  332. return -EINVAL;
  333. }
  334. /* prepare to send now (add options etc.) */
  335. dp->dccps_hc_tx_insert_options = 1;
  336. DCCP_SKB_CB(skb)->dccpd_ccval = hctx->ccid3hctx_last_win_count;
  337. hctx->ccid3hctx_idle = 0;
  338. /* set the nominal send time for the next following packet */
  339. timeval_add_usecs(&hctx->ccid3hctx_t_nom, hctx->ccid3hctx_t_ipi);
  340. return 0;
  341. }
  342. static void ccid3_hc_tx_packet_sent(struct sock *sk, int more,
  343. unsigned int len)
  344. {
  345. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  346. struct timeval now;
  347. struct dccp_tx_hist_entry *packet;
  348. BUG_ON(hctx == NULL);
  349. ccid3_hc_tx_update_s(hctx, len);
  350. packet = dccp_tx_hist_entry_new(ccid3_tx_hist, GFP_ATOMIC);
  351. if (unlikely(packet == NULL)) {
  352. DCCP_CRIT("packet history - out of memory!");
  353. return;
  354. }
  355. dccp_tx_hist_add_entry(&hctx->ccid3hctx_hist, packet);
  356. dccp_timestamp(sk, &now);
  357. packet->dccphtx_tstamp = now;
  358. packet->dccphtx_seqno = dccp_sk(sk)->dccps_gss;
  359. packet->dccphtx_rtt = hctx->ccid3hctx_rtt;
  360. packet->dccphtx_sent = 1;
  361. }
  362. static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
  363. {
  364. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  365. struct ccid3_options_received *opt_recv;
  366. struct dccp_tx_hist_entry *packet;
  367. struct timeval now;
  368. unsigned long t_nfb;
  369. u32 pinv, r_sample;
  370. BUG_ON(hctx == NULL);
  371. /* we are only interested in ACKs */
  372. if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK ||
  373. DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK))
  374. return;
  375. opt_recv = &hctx->ccid3hctx_options_received;
  376. switch (hctx->ccid3hctx_state) {
  377. case TFRC_SSTATE_NO_FBACK:
  378. case TFRC_SSTATE_FBACK:
  379. /* get packet from history to look up t_recvdata */
  380. packet = dccp_tx_hist_find_entry(&hctx->ccid3hctx_hist,
  381. DCCP_SKB_CB(skb)->dccpd_ack_seq);
  382. if (unlikely(packet == NULL)) {
  383. DCCP_WARN("%s(%p), seqno %llu(%s) doesn't exist "
  384. "in history!\n", dccp_role(sk), sk,
  385. (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
  386. dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type));
  387. return;
  388. }
  389. /* Update receive rate in units of 64 * bytes/second */
  390. hctx->ccid3hctx_x_recv = opt_recv->ccid3or_receive_rate;
  391. hctx->ccid3hctx_x_recv <<= 6;
  392. /* Update loss event rate */
  393. pinv = opt_recv->ccid3or_loss_event_rate;
  394. if (pinv == ~0U || pinv == 0) /* see RFC 4342, 8.5 */
  395. hctx->ccid3hctx_p = 0;
  396. else /* can not exceed 100% */
  397. hctx->ccid3hctx_p = 1000000 / pinv;
  398. dccp_timestamp(sk, &now);
  399. /*
  400. * Calculate new round trip sample as per [RFC 3448, 4.3] by
  401. * R_sample = (now - t_recvdata) - t_elapsed
  402. */
  403. r_sample = dccp_sample_rtt(sk, &now, &packet->dccphtx_tstamp);
  404. /*
  405. * Update RTT estimate by
  406. * If (No feedback recv)
  407. * R = R_sample;
  408. * Else
  409. * R = q * R + (1 - q) * R_sample;
  410. *
  411. * q is a constant, RFC 3448 recomments 0.9
  412. */
  413. if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
  414. /*
  415. * Larger Initial Windows [RFC 4342, sec. 5]
  416. */
  417. hctx->ccid3hctx_rtt = r_sample;
  418. hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
  419. hctx->ccid3hctx_t_ld = now;
  420. ccid3_update_send_interval(hctx);
  421. ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
  422. "R_sample=%uus, X=%u\n", dccp_role(sk),
  423. sk, hctx->ccid3hctx_s,
  424. dccp_sk(sk)->dccps_mss_cache, r_sample,
  425. (unsigned)(hctx->ccid3hctx_x >> 6));
  426. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
  427. } else {
  428. hctx->ccid3hctx_rtt = (9 * hctx->ccid3hctx_rtt +
  429. r_sample) / 10;
  430. /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
  431. if (hctx->ccid3hctx_p > 0)
  432. hctx->ccid3hctx_x_calc =
  433. tfrc_calc_x(hctx->ccid3hctx_s,
  434. hctx->ccid3hctx_rtt,
  435. hctx->ccid3hctx_p);
  436. ccid3_hc_tx_update_x(sk, &now);
  437. ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
  438. "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
  439. dccp_role(sk),
  440. sk, hctx->ccid3hctx_rtt, r_sample,
  441. hctx->ccid3hctx_s, hctx->ccid3hctx_p,
  442. hctx->ccid3hctx_x_calc,
  443. (unsigned)(hctx->ccid3hctx_x_recv >> 6),
  444. (unsigned)(hctx->ccid3hctx_x >> 6));
  445. }
  446. /* unschedule no feedback timer */
  447. sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
  448. /* remove all packets older than the one acked from history */
  449. dccp_tx_hist_purge_older(ccid3_tx_hist,
  450. &hctx->ccid3hctx_hist, packet);
  451. /*
  452. * As we have calculated new ipi, delta, t_nom it is possible
  453. * that we now can send a packet, so wake up dccp_wait_for_ccid
  454. */
  455. sk->sk_write_space(sk);
  456. /*
  457. * Update timeout interval for the nofeedback timer.
  458. * We use a configuration option to increase the lower bound.
  459. * This can help avoid triggering the nofeedback timer too
  460. * often ('spinning') on LANs with small RTTs.
  461. */
  462. hctx->ccid3hctx_t_rto = max_t(u32, 4 * hctx->ccid3hctx_rtt,
  463. CONFIG_IP_DCCP_CCID3_RTO *
  464. (USEC_PER_SEC/1000));
  465. /*
  466. * Schedule no feedback timer to expire in
  467. * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
  468. */
  469. t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
  470. ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
  471. "expire in %lu jiffies (%luus)\n",
  472. dccp_role(sk),
  473. sk, usecs_to_jiffies(t_nfb), t_nfb);
  474. sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
  475. jiffies + usecs_to_jiffies(t_nfb));
  476. /* set idle flag */
  477. hctx->ccid3hctx_idle = 1;
  478. break;
  479. case TFRC_SSTATE_NO_SENT: /* fall through */
  480. case TFRC_SSTATE_TERM: /* ignore feedback when closing */
  481. break;
  482. }
  483. }
  484. static int ccid3_hc_tx_parse_options(struct sock *sk, unsigned char option,
  485. unsigned char len, u16 idx,
  486. unsigned char *value)
  487. {
  488. int rc = 0;
  489. const struct dccp_sock *dp = dccp_sk(sk);
  490. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  491. struct ccid3_options_received *opt_recv;
  492. BUG_ON(hctx == NULL);
  493. opt_recv = &hctx->ccid3hctx_options_received;
  494. if (opt_recv->ccid3or_seqno != dp->dccps_gsr) {
  495. opt_recv->ccid3or_seqno = dp->dccps_gsr;
  496. opt_recv->ccid3or_loss_event_rate = ~0;
  497. opt_recv->ccid3or_loss_intervals_idx = 0;
  498. opt_recv->ccid3or_loss_intervals_len = 0;
  499. opt_recv->ccid3or_receive_rate = 0;
  500. }
  501. switch (option) {
  502. case TFRC_OPT_LOSS_EVENT_RATE:
  503. if (unlikely(len != 4)) {
  504. DCCP_WARN("%s(%p), invalid len %d "
  505. "for TFRC_OPT_LOSS_EVENT_RATE\n",
  506. dccp_role(sk), sk, len);
  507. rc = -EINVAL;
  508. } else {
  509. opt_recv->ccid3or_loss_event_rate =
  510. ntohl(*(__be32 *)value);
  511. ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
  512. dccp_role(sk), sk,
  513. opt_recv->ccid3or_loss_event_rate);
  514. }
  515. break;
  516. case TFRC_OPT_LOSS_INTERVALS:
  517. opt_recv->ccid3or_loss_intervals_idx = idx;
  518. opt_recv->ccid3or_loss_intervals_len = len;
  519. ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
  520. dccp_role(sk), sk,
  521. opt_recv->ccid3or_loss_intervals_idx,
  522. opt_recv->ccid3or_loss_intervals_len);
  523. break;
  524. case TFRC_OPT_RECEIVE_RATE:
  525. if (unlikely(len != 4)) {
  526. DCCP_WARN("%s(%p), invalid len %d "
  527. "for TFRC_OPT_RECEIVE_RATE\n",
  528. dccp_role(sk), sk, len);
  529. rc = -EINVAL;
  530. } else {
  531. opt_recv->ccid3or_receive_rate =
  532. ntohl(*(__be32 *)value);
  533. ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
  534. dccp_role(sk), sk,
  535. opt_recv->ccid3or_receive_rate);
  536. }
  537. break;
  538. }
  539. return rc;
  540. }
  541. static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk)
  542. {
  543. struct ccid3_hc_tx_sock *hctx = ccid_priv(ccid);
  544. hctx->ccid3hctx_s = 0;
  545. hctx->ccid3hctx_rtt = 0;
  546. hctx->ccid3hctx_state = TFRC_SSTATE_NO_SENT;
  547. INIT_LIST_HEAD(&hctx->ccid3hctx_hist);
  548. hctx->ccid3hctx_no_feedback_timer.function =
  549. ccid3_hc_tx_no_feedback_timer;
  550. hctx->ccid3hctx_no_feedback_timer.data = (unsigned long)sk;
  551. init_timer(&hctx->ccid3hctx_no_feedback_timer);
  552. return 0;
  553. }
  554. static void ccid3_hc_tx_exit(struct sock *sk)
  555. {
  556. struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  557. BUG_ON(hctx == NULL);
  558. ccid3_hc_tx_set_state(sk, TFRC_SSTATE_TERM);
  559. sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
  560. /* Empty packet history */
  561. dccp_tx_hist_purge(ccid3_tx_hist, &hctx->ccid3hctx_hist);
  562. }
  563. static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info)
  564. {
  565. const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  566. /* Listen socks doesn't have a private CCID block */
  567. if (sk->sk_state == DCCP_LISTEN)
  568. return;
  569. BUG_ON(hctx == NULL);
  570. info->tcpi_rto = hctx->ccid3hctx_t_rto;
  571. info->tcpi_rtt = hctx->ccid3hctx_rtt;
  572. }
  573. static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
  574. u32 __user *optval, int __user *optlen)
  575. {
  576. const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
  577. const void *val;
  578. /* Listen socks doesn't have a private CCID block */
  579. if (sk->sk_state == DCCP_LISTEN)
  580. return -EINVAL;
  581. switch (optname) {
  582. case DCCP_SOCKOPT_CCID_TX_INFO:
  583. if (len < sizeof(hctx->ccid3hctx_tfrc))
  584. return -EINVAL;
  585. len = sizeof(hctx->ccid3hctx_tfrc);
  586. val = &hctx->ccid3hctx_tfrc;
  587. break;
  588. default:
  589. return -ENOPROTOOPT;
  590. }
  591. if (put_user(len, optlen) || copy_to_user(optval, val, len))
  592. return -EFAULT;
  593. return 0;
  594. }
  595. /*
  596. * Receiver Half-Connection Routines
  597. */
  598. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  599. static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state)
  600. {
  601. static char *ccid3_rx_state_names[] = {
  602. [TFRC_RSTATE_NO_DATA] = "NO_DATA",
  603. [TFRC_RSTATE_DATA] = "DATA",
  604. [TFRC_RSTATE_TERM] = "TERM",
  605. };
  606. return ccid3_rx_state_names[state];
  607. }
  608. #endif
  609. static void ccid3_hc_rx_set_state(struct sock *sk,
  610. enum ccid3_hc_rx_states state)
  611. {
  612. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  613. enum ccid3_hc_rx_states oldstate = hcrx->ccid3hcrx_state;
  614. ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
  615. dccp_role(sk), sk, ccid3_rx_state_name(oldstate),
  616. ccid3_rx_state_name(state));
  617. WARN_ON(state == oldstate);
  618. hcrx->ccid3hcrx_state = state;
  619. }
  620. static inline void ccid3_hc_rx_update_s(struct ccid3_hc_rx_sock *hcrx, int len)
  621. {
  622. if (unlikely(len == 0)) /* don't update on empty packets (e.g. ACKs) */
  623. ccid3_pr_debug("Packet payload length is 0 - not updating\n");
  624. else
  625. hcrx->ccid3hcrx_s = hcrx->ccid3hcrx_s == 0 ? len :
  626. (9 * hcrx->ccid3hcrx_s + len) / 10;
  627. }
  628. static void ccid3_hc_rx_send_feedback(struct sock *sk)
  629. {
  630. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  631. struct dccp_sock *dp = dccp_sk(sk);
  632. struct dccp_rx_hist_entry *packet;
  633. struct timeval now;
  634. suseconds_t delta;
  635. ccid3_pr_debug("%s(%p) - entry \n", dccp_role(sk), sk);
  636. dccp_timestamp(sk, &now);
  637. switch (hcrx->ccid3hcrx_state) {
  638. case TFRC_RSTATE_NO_DATA:
  639. hcrx->ccid3hcrx_x_recv = 0;
  640. break;
  641. case TFRC_RSTATE_DATA:
  642. delta = timeval_delta(&now,
  643. &hcrx->ccid3hcrx_tstamp_last_feedback);
  644. DCCP_BUG_ON(delta < 0);
  645. hcrx->ccid3hcrx_x_recv =
  646. scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
  647. break;
  648. case TFRC_RSTATE_TERM:
  649. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  650. return;
  651. }
  652. packet = dccp_rx_hist_find_data_packet(&hcrx->ccid3hcrx_hist);
  653. if (unlikely(packet == NULL)) {
  654. DCCP_WARN("%s(%p), no data packet in history!\n",
  655. dccp_role(sk), sk);
  656. return;
  657. }
  658. hcrx->ccid3hcrx_tstamp_last_feedback = now;
  659. hcrx->ccid3hcrx_ccval_last_counter = packet->dccphrx_ccval;
  660. hcrx->ccid3hcrx_bytes_recv = 0;
  661. /* Elapsed time information [RFC 4340, 13.2] in units of 10 * usecs */
  662. delta = timeval_delta(&now, &packet->dccphrx_tstamp);
  663. DCCP_BUG_ON(delta < 0);
  664. hcrx->ccid3hcrx_elapsed_time = delta / 10;
  665. if (hcrx->ccid3hcrx_p == 0)
  666. hcrx->ccid3hcrx_pinv = ~0U; /* see RFC 4342, 8.5 */
  667. else if (hcrx->ccid3hcrx_p > 1000000) {
  668. DCCP_WARN("p (%u) > 100%%\n", hcrx->ccid3hcrx_p);
  669. hcrx->ccid3hcrx_pinv = 1; /* use 100% in this case */
  670. } else
  671. hcrx->ccid3hcrx_pinv = 1000000 / hcrx->ccid3hcrx_p;
  672. dp->dccps_hc_rx_insert_options = 1;
  673. dccp_send_ack(sk);
  674. }
  675. static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
  676. {
  677. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  678. __be32 x_recv, pinv;
  679. BUG_ON(hcrx == NULL);
  680. if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN))
  681. return 0;
  682. DCCP_SKB_CB(skb)->dccpd_ccval = hcrx->ccid3hcrx_ccval_last_counter;
  683. if (dccp_packet_without_ack(skb))
  684. return 0;
  685. x_recv = htonl(hcrx->ccid3hcrx_x_recv);
  686. pinv = htonl(hcrx->ccid3hcrx_pinv);
  687. if ((hcrx->ccid3hcrx_elapsed_time != 0 &&
  688. dccp_insert_option_elapsed_time(sk, skb,
  689. hcrx->ccid3hcrx_elapsed_time)) ||
  690. dccp_insert_option_timestamp(sk, skb) ||
  691. dccp_insert_option(sk, skb, TFRC_OPT_LOSS_EVENT_RATE,
  692. &pinv, sizeof(pinv)) ||
  693. dccp_insert_option(sk, skb, TFRC_OPT_RECEIVE_RATE,
  694. &x_recv, sizeof(x_recv)))
  695. return -1;
  696. return 0;
  697. }
  698. /* calculate first loss interval
  699. *
  700. * returns estimated loss interval in usecs */
  701. static u32 ccid3_hc_rx_calc_first_li(struct sock *sk)
  702. {
  703. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  704. struct dccp_rx_hist_entry *entry, *next, *tail = NULL;
  705. u32 x_recv, p;
  706. suseconds_t rtt, delta;
  707. struct timeval tstamp = { 0, };
  708. int interval = 0;
  709. int win_count = 0;
  710. int step = 0;
  711. u64 fval;
  712. list_for_each_entry_safe(entry, next, &hcrx->ccid3hcrx_hist,
  713. dccphrx_node) {
  714. if (dccp_rx_hist_entry_data_packet(entry)) {
  715. tail = entry;
  716. switch (step) {
  717. case 0:
  718. tstamp = entry->dccphrx_tstamp;
  719. win_count = entry->dccphrx_ccval;
  720. step = 1;
  721. break;
  722. case 1:
  723. interval = win_count - entry->dccphrx_ccval;
  724. if (interval < 0)
  725. interval += TFRC_WIN_COUNT_LIMIT;
  726. if (interval > 4)
  727. goto found;
  728. break;
  729. }
  730. }
  731. }
  732. if (unlikely(step == 0)) {
  733. DCCP_WARN("%s(%p), packet history has no data packets!\n",
  734. dccp_role(sk), sk);
  735. return ~0;
  736. }
  737. if (unlikely(interval == 0)) {
  738. DCCP_WARN("%s(%p), Could not find a win_count interval > 0."
  739. "Defaulting to 1\n", dccp_role(sk), sk);
  740. interval = 1;
  741. }
  742. found:
  743. if (!tail) {
  744. DCCP_CRIT("tail is null\n");
  745. return ~0;
  746. }
  747. delta = timeval_delta(&tstamp, &tail->dccphrx_tstamp);
  748. DCCP_BUG_ON(delta < 0);
  749. rtt = delta * 4 / interval;
  750. ccid3_pr_debug("%s(%p), approximated RTT to %dus\n",
  751. dccp_role(sk), sk, (int)rtt);
  752. /*
  753. * Determine the length of the first loss interval via inverse lookup.
  754. * Assume that X_recv can be computed by the throughput equation
  755. * s
  756. * X_recv = --------
  757. * R * fval
  758. * Find some p such that f(p) = fval; return 1/p [RFC 3448, 6.3.1].
  759. */
  760. if (rtt == 0) { /* would result in divide-by-zero */
  761. DCCP_WARN("RTT==0\n");
  762. return ~0;
  763. }
  764. dccp_timestamp(sk, &tstamp);
  765. delta = timeval_delta(&tstamp, &hcrx->ccid3hcrx_tstamp_last_feedback);
  766. DCCP_BUG_ON(delta <= 0);
  767. x_recv = scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
  768. if (x_recv == 0) { /* would also trigger divide-by-zero */
  769. DCCP_WARN("X_recv==0\n");
  770. if ((x_recv = hcrx->ccid3hcrx_x_recv) == 0) {
  771. DCCP_BUG("stored value of X_recv is zero");
  772. return ~0;
  773. }
  774. }
  775. fval = scaled_div(hcrx->ccid3hcrx_s, rtt);
  776. fval = scaled_div32(fval, x_recv);
  777. p = tfrc_calc_x_reverse_lookup(fval);
  778. ccid3_pr_debug("%s(%p), receive rate=%u bytes/s, implied "
  779. "loss rate=%u\n", dccp_role(sk), sk, x_recv, p);
  780. if (p == 0)
  781. return ~0;
  782. else
  783. return 1000000 / p;
  784. }
  785. static void ccid3_hc_rx_update_li(struct sock *sk, u64 seq_loss, u8 win_loss)
  786. {
  787. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  788. struct dccp_li_hist_entry *head;
  789. u64 seq_temp;
  790. if (list_empty(&hcrx->ccid3hcrx_li_hist)) {
  791. if (!dccp_li_hist_interval_new(ccid3_li_hist,
  792. &hcrx->ccid3hcrx_li_hist, seq_loss, win_loss))
  793. return;
  794. head = list_entry(hcrx->ccid3hcrx_li_hist.next,
  795. struct dccp_li_hist_entry, dccplih_node);
  796. head->dccplih_interval = ccid3_hc_rx_calc_first_li(sk);
  797. } else {
  798. struct dccp_li_hist_entry *entry;
  799. struct list_head *tail;
  800. head = list_entry(hcrx->ccid3hcrx_li_hist.next,
  801. struct dccp_li_hist_entry, dccplih_node);
  802. /* FIXME win count check removed as was wrong */
  803. /* should make this check with receive history */
  804. /* and compare there as per section 10.2 of RFC4342 */
  805. /* new loss event detected */
  806. /* calculate last interval length */
  807. seq_temp = dccp_delta_seqno(head->dccplih_seqno, seq_loss);
  808. entry = dccp_li_hist_entry_new(ccid3_li_hist, GFP_ATOMIC);
  809. if (entry == NULL) {
  810. DCCP_BUG("out of memory - can not allocate entry");
  811. return;
  812. }
  813. list_add(&entry->dccplih_node, &hcrx->ccid3hcrx_li_hist);
  814. tail = hcrx->ccid3hcrx_li_hist.prev;
  815. list_del(tail);
  816. kmem_cache_free(ccid3_li_hist->dccplih_slab, tail);
  817. /* Create the newest interval */
  818. entry->dccplih_seqno = seq_loss;
  819. entry->dccplih_interval = seq_temp;
  820. entry->dccplih_win_count = win_loss;
  821. }
  822. }
  823. static int ccid3_hc_rx_detect_loss(struct sock *sk,
  824. struct dccp_rx_hist_entry *packet)
  825. {
  826. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  827. struct dccp_rx_hist_entry *rx_hist =
  828. dccp_rx_hist_head(&hcrx->ccid3hcrx_hist);
  829. u64 seqno = packet->dccphrx_seqno;
  830. u64 tmp_seqno;
  831. int loss = 0;
  832. u8 ccval;
  833. tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
  834. if (!rx_hist ||
  835. follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
  836. hcrx->ccid3hcrx_seqno_nonloss = seqno;
  837. hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
  838. goto detect_out;
  839. }
  840. while (dccp_delta_seqno(hcrx->ccid3hcrx_seqno_nonloss, seqno)
  841. > TFRC_RECV_NUM_LATE_LOSS) {
  842. loss = 1;
  843. ccid3_hc_rx_update_li(sk, hcrx->ccid3hcrx_seqno_nonloss,
  844. hcrx->ccid3hcrx_ccval_nonloss);
  845. tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
  846. dccp_inc_seqno(&tmp_seqno);
  847. hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
  848. dccp_inc_seqno(&tmp_seqno);
  849. while (dccp_rx_hist_find_entry(&hcrx->ccid3hcrx_hist,
  850. tmp_seqno, &ccval)) {
  851. hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
  852. hcrx->ccid3hcrx_ccval_nonloss = ccval;
  853. dccp_inc_seqno(&tmp_seqno);
  854. }
  855. }
  856. /* FIXME - this code could be simplified with above while */
  857. /* but works at moment */
  858. if (follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
  859. hcrx->ccid3hcrx_seqno_nonloss = seqno;
  860. hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
  861. }
  862. detect_out:
  863. dccp_rx_hist_add_packet(ccid3_rx_hist, &hcrx->ccid3hcrx_hist,
  864. &hcrx->ccid3hcrx_li_hist, packet,
  865. hcrx->ccid3hcrx_seqno_nonloss);
  866. return loss;
  867. }
  868. static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
  869. {
  870. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  871. const struct dccp_options_received *opt_recv;
  872. struct dccp_rx_hist_entry *packet;
  873. struct timeval now;
  874. u32 p_prev, r_sample, rtt_prev;
  875. int loss, payload_size;
  876. BUG_ON(hcrx == NULL);
  877. opt_recv = &dccp_sk(sk)->dccps_options_received;
  878. switch (DCCP_SKB_CB(skb)->dccpd_type) {
  879. case DCCP_PKT_ACK:
  880. if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
  881. return;
  882. case DCCP_PKT_DATAACK:
  883. if (opt_recv->dccpor_timestamp_echo == 0)
  884. break;
  885. rtt_prev = hcrx->ccid3hcrx_rtt;
  886. dccp_timestamp(sk, &now);
  887. r_sample = dccp_sample_rtt(sk, &now, NULL);
  888. if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
  889. hcrx->ccid3hcrx_rtt = r_sample;
  890. else
  891. hcrx->ccid3hcrx_rtt = (hcrx->ccid3hcrx_rtt * 9) / 10 +
  892. r_sample / 10;
  893. if (rtt_prev != hcrx->ccid3hcrx_rtt)
  894. ccid3_pr_debug("%s(%p), New RTT=%uus, elapsed time=%u\n",
  895. dccp_role(sk), sk, hcrx->ccid3hcrx_rtt,
  896. opt_recv->dccpor_elapsed_time);
  897. break;
  898. case DCCP_PKT_DATA:
  899. break;
  900. default: /* We're not interested in other packet types, move along */
  901. return;
  902. }
  903. packet = dccp_rx_hist_entry_new(ccid3_rx_hist, sk, opt_recv->dccpor_ndp,
  904. skb, GFP_ATOMIC);
  905. if (unlikely(packet == NULL)) {
  906. DCCP_WARN("%s(%p), Not enough mem to add rx packet "
  907. "to history, consider it lost!\n", dccp_role(sk), sk);
  908. return;
  909. }
  910. loss = ccid3_hc_rx_detect_loss(sk, packet);
  911. if (DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK)
  912. return;
  913. payload_size = skb->len - dccp_hdr(skb)->dccph_doff * 4;
  914. ccid3_hc_rx_update_s(hcrx, payload_size);
  915. switch (hcrx->ccid3hcrx_state) {
  916. case TFRC_RSTATE_NO_DATA:
  917. ccid3_pr_debug("%s(%p, state=%s), skb=%p, sending initial "
  918. "feedback\n", dccp_role(sk), sk,
  919. dccp_state_name(sk->sk_state), skb);
  920. ccid3_hc_rx_send_feedback(sk);
  921. ccid3_hc_rx_set_state(sk, TFRC_RSTATE_DATA);
  922. return;
  923. case TFRC_RSTATE_DATA:
  924. hcrx->ccid3hcrx_bytes_recv += payload_size;
  925. if (loss)
  926. break;
  927. dccp_timestamp(sk, &now);
  928. if ((timeval_delta(&now, &hcrx->ccid3hcrx_tstamp_last_ack) -
  929. (suseconds_t)hcrx->ccid3hcrx_rtt) >= 0) {
  930. hcrx->ccid3hcrx_tstamp_last_ack = now;
  931. ccid3_hc_rx_send_feedback(sk);
  932. }
  933. return;
  934. case TFRC_RSTATE_TERM:
  935. DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
  936. return;
  937. }
  938. /* Dealing with packet loss */
  939. ccid3_pr_debug("%s(%p, state=%s), data loss! Reacting...\n",
  940. dccp_role(sk), sk, dccp_state_name(sk->sk_state));
  941. p_prev = hcrx->ccid3hcrx_p;
  942. /* Calculate loss event rate */
  943. if (!list_empty(&hcrx->ccid3hcrx_li_hist)) {
  944. u32 i_mean = dccp_li_hist_calc_i_mean(&hcrx->ccid3hcrx_li_hist);
  945. /* Scaling up by 1000000 as fixed decimal */
  946. if (i_mean != 0)
  947. hcrx->ccid3hcrx_p = 1000000 / i_mean;
  948. } else
  949. DCCP_BUG("empty loss history");
  950. if (hcrx->ccid3hcrx_p > p_prev) {
  951. ccid3_hc_rx_send_feedback(sk);
  952. return;
  953. }
  954. }
  955. static int ccid3_hc_rx_init(struct ccid *ccid, struct sock *sk)
  956. {
  957. struct ccid3_hc_rx_sock *hcrx = ccid_priv(ccid);
  958. ccid3_pr_debug("entry\n");
  959. hcrx->ccid3hcrx_state = TFRC_RSTATE_NO_DATA;
  960. INIT_LIST_HEAD(&hcrx->ccid3hcrx_hist);
  961. INIT_LIST_HEAD(&hcrx->ccid3hcrx_li_hist);
  962. dccp_timestamp(sk, &hcrx->ccid3hcrx_tstamp_last_ack);
  963. hcrx->ccid3hcrx_tstamp_last_feedback = hcrx->ccid3hcrx_tstamp_last_ack;
  964. hcrx->ccid3hcrx_s = 0;
  965. hcrx->ccid3hcrx_rtt = 0;
  966. return 0;
  967. }
  968. static void ccid3_hc_rx_exit(struct sock *sk)
  969. {
  970. struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  971. BUG_ON(hcrx == NULL);
  972. ccid3_hc_rx_set_state(sk, TFRC_RSTATE_TERM);
  973. /* Empty packet history */
  974. dccp_rx_hist_purge(ccid3_rx_hist, &hcrx->ccid3hcrx_hist);
  975. /* Empty loss interval history */
  976. dccp_li_hist_purge(ccid3_li_hist, &hcrx->ccid3hcrx_li_hist);
  977. }
  978. static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info)
  979. {
  980. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  981. /* Listen socks doesn't have a private CCID block */
  982. if (sk->sk_state == DCCP_LISTEN)
  983. return;
  984. BUG_ON(hcrx == NULL);
  985. info->tcpi_ca_state = hcrx->ccid3hcrx_state;
  986. info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
  987. info->tcpi_rcv_rtt = hcrx->ccid3hcrx_rtt;
  988. }
  989. static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len,
  990. u32 __user *optval, int __user *optlen)
  991. {
  992. const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
  993. const void *val;
  994. /* Listen socks doesn't have a private CCID block */
  995. if (sk->sk_state == DCCP_LISTEN)
  996. return -EINVAL;
  997. switch (optname) {
  998. case DCCP_SOCKOPT_CCID_RX_INFO:
  999. if (len < sizeof(hcrx->ccid3hcrx_tfrc))
  1000. return -EINVAL;
  1001. len = sizeof(hcrx->ccid3hcrx_tfrc);
  1002. val = &hcrx->ccid3hcrx_tfrc;
  1003. break;
  1004. default:
  1005. return -ENOPROTOOPT;
  1006. }
  1007. if (put_user(len, optlen) || copy_to_user(optval, val, len))
  1008. return -EFAULT;
  1009. return 0;
  1010. }
  1011. static struct ccid_operations ccid3 = {
  1012. .ccid_id = DCCPC_CCID3,
  1013. .ccid_name = "ccid3",
  1014. .ccid_owner = THIS_MODULE,
  1015. .ccid_hc_tx_obj_size = sizeof(struct ccid3_hc_tx_sock),
  1016. .ccid_hc_tx_init = ccid3_hc_tx_init,
  1017. .ccid_hc_tx_exit = ccid3_hc_tx_exit,
  1018. .ccid_hc_tx_send_packet = ccid3_hc_tx_send_packet,
  1019. .ccid_hc_tx_packet_sent = ccid3_hc_tx_packet_sent,
  1020. .ccid_hc_tx_packet_recv = ccid3_hc_tx_packet_recv,
  1021. .ccid_hc_tx_parse_options = ccid3_hc_tx_parse_options,
  1022. .ccid_hc_rx_obj_size = sizeof(struct ccid3_hc_rx_sock),
  1023. .ccid_hc_rx_init = ccid3_hc_rx_init,
  1024. .ccid_hc_rx_exit = ccid3_hc_rx_exit,
  1025. .ccid_hc_rx_insert_options = ccid3_hc_rx_insert_options,
  1026. .ccid_hc_rx_packet_recv = ccid3_hc_rx_packet_recv,
  1027. .ccid_hc_rx_get_info = ccid3_hc_rx_get_info,
  1028. .ccid_hc_tx_get_info = ccid3_hc_tx_get_info,
  1029. .ccid_hc_rx_getsockopt = ccid3_hc_rx_getsockopt,
  1030. .ccid_hc_tx_getsockopt = ccid3_hc_tx_getsockopt,
  1031. };
  1032. #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
  1033. module_param(ccid3_debug, int, 0444);
  1034. MODULE_PARM_DESC(ccid3_debug, "Enable debug messages");
  1035. #endif
  1036. static __init int ccid3_module_init(void)
  1037. {
  1038. int rc = -ENOBUFS;
  1039. ccid3_rx_hist = dccp_rx_hist_new("ccid3");
  1040. if (ccid3_rx_hist == NULL)
  1041. goto out;
  1042. ccid3_tx_hist = dccp_tx_hist_new("ccid3");
  1043. if (ccid3_tx_hist == NULL)
  1044. goto out_free_rx;
  1045. ccid3_li_hist = dccp_li_hist_new("ccid3");
  1046. if (ccid3_li_hist == NULL)
  1047. goto out_free_tx;
  1048. rc = ccid_register(&ccid3);
  1049. if (rc != 0)
  1050. goto out_free_loss_interval_history;
  1051. out:
  1052. return rc;
  1053. out_free_loss_interval_history:
  1054. dccp_li_hist_delete(ccid3_li_hist);
  1055. ccid3_li_hist = NULL;
  1056. out_free_tx:
  1057. dccp_tx_hist_delete(ccid3_tx_hist);
  1058. ccid3_tx_hist = NULL;
  1059. out_free_rx:
  1060. dccp_rx_hist_delete(ccid3_rx_hist);
  1061. ccid3_rx_hist = NULL;
  1062. goto out;
  1063. }
  1064. module_init(ccid3_module_init);
  1065. static __exit void ccid3_module_exit(void)
  1066. {
  1067. ccid_unregister(&ccid3);
  1068. if (ccid3_tx_hist != NULL) {
  1069. dccp_tx_hist_delete(ccid3_tx_hist);
  1070. ccid3_tx_hist = NULL;
  1071. }
  1072. if (ccid3_rx_hist != NULL) {
  1073. dccp_rx_hist_delete(ccid3_rx_hist);
  1074. ccid3_rx_hist = NULL;
  1075. }
  1076. if (ccid3_li_hist != NULL) {
  1077. dccp_li_hist_delete(ccid3_li_hist);
  1078. ccid3_li_hist = NULL;
  1079. }
  1080. }
  1081. module_exit(ccid3_module_exit);
  1082. MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
  1083. "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
  1084. MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
  1085. MODULE_LICENSE("GPL");
  1086. MODULE_ALIAS("net-dccp-ccid-3");